The only A/B testing tool that won't slow down your website performance. Get started for free.
Get started for FREE
No-code. Free forever.
Min read

How Long Should I Run My A/B Test?

Donald Ng
October 7, 2023
5-star rating
Reviews on Capterra

A/B testing: Love it or fear it, it’s a game-changer for any marketer.

It could be a shot in the dark or the arrow that hits the bullseye on your conversion rates, all depending on how you execute it. I’ve seen how remarkable the results of an A/B test can be, and trust me; the numbers don’t lie. But, the million-dollar question that always seems to hover like a storm cloud on a sunny day is, “How long should I run my A/B test?”

It’s a valid question, and, indeed, the answer isn’t as apparent as it might seem. Hence, we sat down, dug into the data, and we’re here to answer this question for you.

So let’s get started!

Does A/B Testing Really Boost Conversion Rates?

In simple terms, an A/B test compares two versions of the same concept to see which one performs better. Although it’s been around for almost 100 years, many marketers still shy away from running A/B tests. The main reason? Most of them don’t know how long they should run an A/B test or how to set them up for accurate results.

To demonstrate the effectiveness of A/B testing, let’s take a look at a fascinating experiment by Google, famously known as the ‘50 Shades of Blue’ test. This peculiarly named test was designed to answer the question: Can the color of links in the Google Search Ads affect the click-through rates?

A team at Google, led by Marissa Meyer, who was the head of product at the time, ran a series of A/B tests. They presented different shades of blue to 1% of their users at a time, running over 40 experiments with all conceivable shades of blue.

In the end, a particular purplish shade of blue emerged as the winner. The test proved to be far from pointless: Google reportedly saw a staggering $200 million increase in revenue attributed to this minor colour change!

Isn’t that astonishing? A subtle alteration in the colour shade made such a significant difference in user click behaviour. It shows every little detail counts when it comes to A/B testing and user experience.

Aside from this, several success stories prove the effectiveness of A/B testing. For instance, increased its conversions by 11.5% just by adding some FAQs, statistics, and social proof.

Then the question arises - how long should you conduct your A/B testing?

How Long Should You Run Your A/B Test?

The simple answer is, run it as long as it takes to reach a statistical significance of 95%-99%. That means, you can be 95%-99% confident that your outcome is valid.

Rushing your A/B test and stopping it prematurely could sway the results, rendering the test ineffective. Although it may be tempting to stop when you get the results you wanted, you should always strive for a high statistical significance.

As a rule of thumb, run your tests for at least two weeks. Web traffic and conversions fluctuate, so testing for a full week will give you a complete picture. If you’re still not at 95% significance, test for another week.

Size Matters

The result of your A/B test is also influenced by the size of your sample. If it’s too small, your margin of error will increase. Think of it as drawing jellybeans from a bag. If you only draw a few and base your entire conclusion on them, it’s likely to be inaccurate.

Generally, you should have at least 1,000 subjects (or conversions, customers, visitors, etc.) in your test. To make sure your sample represents your entire user base, consider running separate tests for different devices and browsers.


Patience truly is a virtue when it comes to A/B testing. And sometimes, if there’s no clear winner after months of testing, it might be time to start again with a new set of variants. Remember, the goal is to gain valuable insights, not just to get it over with quickly.

To sum it up, the key to effective A/B testing is to form the right hypothesis, aim for a high statistical significance, ensure a large enough sample size, and give your test enough time. In other words, be precise and patient.

A/B testing platform for people who
care about  website performance

Mida is 10X faster than everything you have ever considered. Try it yourself.